Type 2 diabetes mellitus is increasing in epidemic proportions worldwide. Control of diabetes requires an understanding of the genes and gene-environment interactions involved in its pathogenesis. Linkage analysis is a key initial strategy to identify genes for complex disorders like diabetes, but many studies provide only suggestive evidence for genomic linkage to type 2 diabetes or related traits. The heterogeneous type 2 diabetes phenotype contributes to modest linkage signals. In particular, three major phenotypic diabetes risk factors: parental diabetes, offspring obesity, and older age-related onset, all introduce heterogeneity that weakens the association of genes with expression of diabetes. The Framingham Heart Study (FHS) has collected extensive cross-sectional and longitudinal phenotypic data and typed genomic micro satellite markers on 1702 parents and offspring within 330 pedigrees from a relatively homogeneous community. In this R21 application we propose secondary analyses of existing FHS data as a cost-effective means to test hypotheses about diabetes genetics and to guide next steps for positional cloning. We will focus on chromosomes Iq, lOq, and 1 Iq, where there is suggestive linkage to diabetes traits in unstratified analyses. We will reduce phenotypic heterogeneity by defining phenotypic sub-strata, including families with and without paternal and/or maternal diabetes, relatively lean vs. relatively obese families, or families with relatively younger vs. relatively older-onset diabetes. Diabetes phenotypes include incident diabetes or diabetes-related quantitative traits (plasma levels of HbAic, glucose and insulin). Analyses will use variance components (VC) models of ordered subsets of strata, or will include formal tests of gene-by-strata interaction, or account for imprinting effects, variable age-of-onset phenotypes (in survival analyses), or for time-varying traits (in VC longitudinal trait analyses). Our hypothesis is that linkage analyses that reduce heterogeneity will refine evidence for linkage on Iq, 10q, 1 Iq, and will identify sub-strata where diabetes genes are more likely to be segregating. Results will provide a foundation for future laboratory efforts and will increase understanding of phenotype-genotype interactions in diabetes pathogenesis.